ndimiduk commented on a change in pull request #830: HBASE-23281: Track meta region locations in masters URL: https://github.com/apache/hbase/pull/830#discussion_r349376520
########## File path: hbase-server/src/main/java/org/apache/hadoop/hbase/master/MetaRegionLocationCache.java ########## @@ -0,0 +1,211 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package org.apache.hadoop.hbase.master; + +import java.util.ArrayList; +import java.util.List; +import java.util.Optional; +import java.util.concurrent.ConcurrentNavigableMap; +import org.apache.hadoop.hbase.HRegionLocation; +import org.apache.hadoop.hbase.exceptions.DeserializationException; +import org.apache.hadoop.hbase.types.CopyOnWriteArrayMap; +import org.apache.hadoop.hbase.util.RetryCounter; +import org.apache.hadoop.hbase.util.RetryCounterFactory; +import org.apache.hadoop.hbase.zookeeper.ZKListener; +import org.apache.hadoop.hbase.zookeeper.ZKUtil; +import org.apache.hadoop.hbase.zookeeper.ZKWatcher; +import org.apache.hadoop.hbase.zookeeper.ZNodePaths; +import org.apache.yetus.audience.InterfaceAudience; +import org.apache.zookeeper.KeeperException; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; +import org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil; + +/** + * A cache of meta region location metadata. Registers a listener on ZK to track changes to the + * meta table znodes. Clients are expected to retry if the meta information is stale. This class + * is thread-safe. + */ [email protected] +public class MetaRegionLocationCache extends ZKListener { + + private static final Logger LOG = LoggerFactory.getLogger(MetaRegionLocationCache.class); + + // Maximum number of times we retry when ZK operation times out. + private static final int MAX_ZK_META_FETCH_RETRIES = 10; + // Sleep interval ms between ZK operation retries. + private static final int SLEEP_INTERVAL_MS_BETWEEN_RETRIES = 1000; + private final RetryCounterFactory retryCounterFactory = + new RetryCounterFactory(MAX_ZK_META_FETCH_RETRIES, SLEEP_INTERVAL_MS_BETWEEN_RETRIES); + + // Cached meta region locations indexed by replica ID. + // CopyOnWriteArrayMap ensures synchronization during updates and a consistent snapshot during + // client requests. Even though CopyOnWriteArrayMap copies the data structure for every write, + // that should be OK since the size of the list is often small and mutations are not too often + // and we do not need to block client requests while mutations are in progress. + private final CopyOnWriteArrayMap<Integer, HRegionLocation> cachedMetaLocations; + + private enum ZNodeOpType { + INIT, + CREATED, + CHANGED, + DELETED + }; + + MetaRegionLocationCache(ZKWatcher zkWatcher) { + super(zkWatcher); + cachedMetaLocations = new CopyOnWriteArrayMap<>(); + watcher.registerListener(this); + // Populate the initial snapshot of data from meta znodes. + // This is needed because stand-by masters can potentially start after the initial znode Review comment: Yeah I think it's more clear. One more improvement would be to distinguish in your method names whether a method is handling the parent znode or one of the replicaId children. I was misunderstanding how watch notification works -- specifically I was expecting that a watch notification event for `nodeDataChanged` includes the new node data and it's updated version information; not the case. I now understand why you have all this back-and-forth with zk. Writing a good test for these situations is difficult without inserting barriers into the masters. Let me think about how to do that in our test harness. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
